Deep Learning based Robust Tomato Plant Leaf Disease Detection and Classification Model on Agricultural Sector

Abstract

Tomato is an indispensable and edible crop across the globe. Tomatoes can be varied in quantity based on how it is fertilized. The main element affecting the quality and quantity of crop yield is Leaf disease. Initial detection of diseases would minimize the disease’s effect on tomato plants and have good crop yield. Various new methods of classifying and identifying some diseases were leveraged broadly. In recent times, many authors (universities, institutes, and labs) have examined and formulated several conventional deep learning (DL) and machine learning (ML) techniques for classification of plant diseases. Therefore, this article introduces a Deep Learning related Robust Tomato Plant Leaf Disease Detection and Classification (DL-TPLFDC) model. To detect tomato plant leaf diseases, the DL-TPLFDC technique initially executes U-Net segmentation technique to identify the leaf portions in the input image. Next, the fuzzy c-means (FCM) clustering with customized binary thresholding process takes place to detect the diseased leaf portions in the preprocessed image. Moreover, the DenseNet-169 model was employed to generate feature vectors from the segmented image. Furthermore, deep variational autoencoder (DVAE) model is applied to eliminate the noisy features exist from the DenseNet-169 model, and the resultant features are fed into the random forest (RF) model for classification process. The design of customized FCM with DVAE based noisy feature removal process demonstrates the novelty of the work. The performance analysis of the DL-TPLFDC technique can be performed on benchmark datasets and the outcomes were examined in various measures. The experimental values portrayed the improved outcomes of the DL-TPLFDC technique over other models.